🔍 Read the full analysis: Google’s Genie 4 Turns AI World-Building Into An Experience on ThorstenMeyerAI.com
Get the latest gadgets delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
Google DeepMind demonstrated Genie 4 on October 27, describing it as a real-time world model that responds to movement and text while generating visuals and audio. The system is not yet publicly available, and DeepMind says worlds can drift over time and support only a few minutes of interaction.
Google DeepMind demonstrated Genie 4 on October 27, presenting an AI system that generates a scene as a user moves through it and gives it text instructions. The company says the model can produce visuals and audio together, but it is not yet generally available, and DeepMind acknowledges that scenes may drift during interactions lasting only a few minutes.
In the demonstrations, users steer generated scenes with keyboard or mouse controls and can type instructions to change elements such as weather or add people and objects. DeepMind describes Genie 4 as a world model: rather than rendering a fixed clip, it predicts what should appear next based on the existing scene and the user’s input. The company says the model does not rely on a stored 3D map like a conventional game engine.
DeepMind also says Genie 4 generates sound alongside images, including ambient audio, music and character dialogue. Its announcement describes Gemini 3.8 Flash as a guiding system that interprets goals, breaks them into steps and monitors generated frames. These are company descriptions of the system; the demonstrations were selected by DeepMind, and the company did not publish benchmark results in the supplied announcement material.
For now, access is limited to trusted testers. DeepMind says it plans to make Genie 4 available through Project Genie to Google AI Ultra subscribers and to early Gemini API partners, but it has not given a launch date or API pricing. DeepMind says every video and audio output carries an invisible SynthID watermark.
Infographic 01 · How Genie grew up
From flat sketches to worlds with sound, in under three years.
Each step as Google DeepMind described it. Session limits and consistency are different measures, so read the labels.
- Feb 2024
Genie 1Simple 2D worlds11 billion parameters, learned from unlabelled internet videos - Dec 2024
Genie 23D worlds from a single imageConsistent “for up to a minute”; real time only as a slimmed-down, lower-quality version - Aug 2025
Genie 3First real-time Genie720p at 24 frames per second, consistent “for a few minutes”, memory about one minute back - Jan 2026
Project GenieUS AI Ultra subscribers can play60 seconds per world, partly because of compute costs - May 2026Project Genie goes globalIn the $200 AI Ultra tier, for users 18 and older
- 27 Oct 2026
Genie 4Sound, live steering, more realismResearch preview for trusted testers; Ultra subscribers and API partners “coming soon”
Genie 3 vs. Genie 4
As described by Google DeepMind. Genie 4 column = vendor claims unless marked.
| Feature | Genie 3 (Aug 2025) | Genie 4 (Oct 2026) |
|---|---|---|
| Sound | Genie 3: None | Genie 4: Generated with the picture: ambient noise, music, dialogue |
| Steering | Genie 3: Keys, plus “promptable world events” | Genie 4: Free-form text changes while it runs; Gemini adapts and checks them |
| Look | Genie 3: 720p, 24 fps | Genie 4: “Unprecedented realism”; no resolution or frame rate published |
| How long | Genie 3: “A few minutes”; Project Genie capped at 60 s | Genie 4: “A few minutes” (DeepMind); up to ~2 min consistent scene reported |
| Access | Genie 3: Small preview, then Project Genie | Genie 4: Trusted testers; Ultra and API partners “coming soon” |
Sources: Genie paper, arXiv, 23 Feb 2024; Genie 2 blog, 4 Dec 2024; Genie 3 blog, 5 Aug 2025; TechCrunch, 29 Jan 2026; Google, 19 May 2026; Google DeepMind, Genie 4 blog, 27 Oct 2026; Genie 4 model page (limitations, disclaimer); TechCrunch, 27 Oct 2026. Table wording: Thorsten Meyer.
Powered by Thorsten Meyer AI
Interactive Worlds Beyond Game Engines
Genie 4 points toward a different way of making interactive digital environments: users can describe changes while a scene is running, rather than building each environment in a traditional game engine or requesting a finished video. If the system becomes reliable and accessible, that approach could be relevant to game prototyping, film previsualization and other work that uses simulated settings.
Those applications are not yet established results. DeepMind reports that testers have used the system for film previsualization and game prototypes, while its robotics example is a research demonstration, not evidence that a physical robot trained or acted in the generated scene. The company’s predecessor, Genie 3, is used in Waymo’s driving simulator, according to the source material, but that does not establish how Genie 4 will perform in commercial or safety-critical settings.
The technology also complicates judgments about synthetic media. A generated scene can include coordinated sound and respond to user input, making it more immersive than a fixed image or clip. DeepMind says SynthID is embedded in outputs, but the announcement does not explain how viewers or platforms can identify that mark in practice.
Products Worth Considering
Logitech MX Master 4 for Mac, Ergonomic Bluetooth Mouse with Haptic Feedback, Ultra-Fast Scrolling, USB-C Charging, for MacOS, iPadOS, Windo
Bose QuietComfort Wireless Bluetooth Headphones with Active Noise Cancellation, Mic, USB‑C Charging, Deep Bass, 24‑Hour Battery – White Smok
LG UltraGear 27-Inch QHD OLED Gaming Monitor, 240Hz, 0.03ms, G-SYNC Compatible, FreeSync Premium Pro, HDMI 2.1
As an affiliate, we earn on qualifying purchases.
How Genie Reached Its Fourth Version
DeepMind’s earlier Genie systems developed in stages. The first Genie, introduced in February 2024, generated simple 2D worlds from video training data. Genie 2 followed in December 2024, creating 3D environments from a single image and maintaining consistency for up to a minute, though real-time operation required a lower-quality version.
In August 2025, Genie 3 became DeepMind’s first world model to run in real time, according to the company. It operated at 720p and 24 frames per second and could maintain a scene for a few minutes, with visual memory reaching back about a minute. Google later made Genie 3 available to paying US subscribers through Project Genie, with TechCrunch reporting a 60-second limit per world. That reported limit was attributed partly to computing costs; it should not be treated as a confirmed Genie 4 limit.
Genie 4’s announcement builds on this progression by emphasizing more realistic scenes, generated audio and text-based changes during interaction. The material provided does not include independent testing or comparable performance measurements against other systems.
““a few minutes of continuous interaction””
— Google DeepMind
Access and Reliability Questions
Several practical details remain open. DeepMind has not announced when Genie 4 will reach Project Genie subscribers or API partners, how much API access will cost, or whether access will be limited by geography, usage or hardware demand. The company also has not published benchmark numbers in the supplied materials, making it difficult to compare the model’s speed, visual consistency or audio quality with alternatives.
The demonstrations do not establish how well the system performs beyond the selected examples. DeepMind itself says objects and details may drift and describes interaction as lasting a few minutes. Reports by TechCrunch and Bloomberg, as cited by heise, put demonstrated scenes at about two minutes, but that duration is not stated in DeepMind’s own announcement. The gap between a demonstration and repeatable performance for users remains untested publicly.
It is also unclear how widely the reported use cases will extend. A simulated robot task does not show that Genie 4 has been used to train a real robot, and tester feedback about film or game workflows does not establish production readiness. The announcement provides no independent evaluation of SynthID’s detectability or of how the watermark will be handled by third-party services.
Awaiting Tester and API Details
The next milestone is the start of access for Google AI Ultra subscribers through Project Genie and early Gemini API partners, which DeepMind has described as coming soon. The company has not supplied a date, price for API use or broader public-release schedule. Those details will determine who can test the system and whether outside developers can build on it.
When access opens, independent users will be able to assess whether the demonstrations translate into stable sessions, how well text instructions affect a live scene and whether audio remains synchronized with events. Until then, Genie 4 is best understood as a company-announced system with promising demonstrated features and acknowledged limits, rather than a broadly available tool with independently verified performance.
Key Questions
What is Google Genie 4?
Genie 4 is a Google DeepMind world model that generates an interactive scene as a user moves through it or provides text instructions. DeepMind says it creates sound along with visuals.
Can the public use Genie 4 now?
No public release is described yet. DeepMind says access is coming soon for trusted testers, Google AI Ultra subscribers through Project Genie and early Gemini API partners. It has not announced a date.
How long can a Genie 4 scene run?
DeepMind says the model supports a few minutes of continuous interaction and warns that details can drift. TechCrunch and Bloomberg, as cited by heise, reported scenes of about two minutes; that figure is not in DeepMind’s own announcement.
Does Genie 4 train physical robots?
The reported robotics example is a simulation involving a virtual Apptronik Apollo humanoid playing checkers. The supplied material says no physical robot was involved, so the demonstration does not show real-world robot training or operation.
How can people identify Genie 4-generated media?
DeepMind says every Genie 4 video and audio output carries an invisible SynthID watermark. The announcement does not specify how viewers can check for it or how third-party platforms will display or detect it.
Source: ThorstenMeyerAI.com
